San Francisco Bay Area
For 10+ years in the industry, I have built production-grade 3D perception systems—spanning geometry-first estimation (SLAM/SfM) and modern neural 3D representations—taking ideas from research prototypes to reliable deployment. My approach is grounded in first principles: spatio-temporally consistent representations, rigorous estimation/optimization, and disciplined systems engineering. I’m particularly interested in how these foundations evolve into the backbone of the future Spatial AI systems. I’ve served as a tech lead / core engineer on high-impact products and pipelines, including Quest SLAM systems, sparse reconstruction for Hyperscape (3DGS), and large-scale crowd-sourced mapping with Mapillary data. I have 30+ publications and U.S./international patents in 3D vision / robotics . Keywords: 3D vision, SLAM, state estimation, sensor fusion, SfM, mapping, 3D reconstruction, 3DGS.
AI for Optimus, FSD etc.
With Meta Reality Labs, I have played roles as - TL and core developer of SLAM mapping system for Quest 3/3S, to enable real-time, high accurate 6DoF tracking on millions of devices: optimization backbone, drift reduction, map streaming, map merge and management. - TL and core developer of sparse reconstruction system for Hyperscape, Meta's Gaussian Splats capture and generation pipeline, released in Connect 2025: SOTA in reconstruction quality and speed - TL and core developer of Meta world mapping system with the data from Mapillary, a Meta owned crowd-source mapping platform, to cover 100 cities, ~5000 km^2 - I work closely with Johannes Schönberger, and contribute to COLMAP internally and externally
Promoted to the head of robotics at Alibaba Cloud/DAMO Academy, directly reporting to Mr. Cheng Li, the CTO of Alibaba Group. - I build and lead a full-stack robot product team of 60 people (45 FTE), including software/hardware engineers, designers, product managers, TPM, etc. Just in 2021, our team have released three new robot products to the market. – One is a suite to build excellent robots, ModuleOne, for mobile robots operating in complicated scenarios, which is composed of one 3D LiDAR, one computing box, and our world-renowned robotic mapping, localization, perception, and navigation software. For ModuleOne, we reached a deal worth $20 Million with the robotics subsidiary of State Grid, which is the largest electricity company on this planet (Fortune 500 Ranking 2nd) on 2022. I represented Alibaba in this collaboration and negotiation. – Another one is the inspection robot, UltraWatt, for industrial applications, e.g. the routine inspections required every day in a power station or chemical plant. – The last one is the community delivery robot, AntMan, which can not only operate outdoor but also take elevators to reach each room in an apartment building to deliver food or packages.
- Under Alibaba AI Labs, I built a world class SLAM team from scratch. The responsibility of my team is to deliver multi-sensor calibration, localization, mapping, and SLAM technology for service robots and automotive use cases. I am also in charge of the robotic perception system design, software architecture and integration. Besides, I collaborated with the hardware team intensively on the sensor synchronization design, sensory data verification, etc. – Delivered a visual-laser localization and mapping technology to our indoor and outdoor service robot products. The sensor suite includes a stereo camera, IMU, odometry, and 2D Lidar (NO RTK GPS). The localization accuracy is 0.05m indoor and 0.30m outdoor. – Productization of our calibration and mapping technologies into cloud services: Calibration-as-a-Service (CaaS)and Mapping-as-a-Service (MaaS), and open to collaborators and customers. – Prototyped a VI-SLAM solution for AR/VR. Our VIO is up to 0.2% drift (stereo, global shutter camera). The localization error of our VI-SLAM is 24% lower than that of VINS-Mono which is the open source SOTA while the computation is 32.5% lower. – Prototyped a Semantic Mapping and Localization system for Automotive use cases. – Published 10+ research papers on top journals/venues.
Working on the next generation of Vision-Inertial SLAM for AR. Granted a US patent “Continuous device pose representation in AR/VR display”.
Do Research & Development of computer vision/sensor fusion/navigation/perception on VR device / Autonomous Driving / Robots.